Method for estimating at least one power consumed by a subset

Abstract

The invention relates to a method for determining the electrical power consumed by a subset of a group (G) comprising at least one subset, each subset of the group (G) comprising at least one electrical device from a set (E) of electrical devices during a period comprising a plurality of elementary periods, the method being implemented by computer and comprising: - estimating the electrical power consumed by the subset by using: - individual elementary operating durations of each subset of the group (G) during the respective elementary periods; and - elementary total energy consumptions of the set (E) during the respective elementary periods.

Description

[0001]Method for estimating at least one power consumed by a subset Field of the invention The field of the invention is that of monitoring the power consumed by a set of equipment. One aim of this type of monitoring is to identify power drifts in order to be able to act on these drifts to improve the energy efficiency of the set of equipment. The invention applies to any industrial installation comprising several energy-consuming equipment and whose share of energy consumption is comparable between equipment. For example, the invention can be applied to drinking water production units, waste sorting, recovery and transformation units and wastewater treatment plants also called STEP in this patent application.For example, most wastewater treatment plants are currently equipped with electricity meters that allow energy consumption to be monitored globally at the plant level, i.e., across all the plant's equipment. In order to identify the most consuming equipment and detect power drifts in the respective equipment, it is necessary to be able to monitor the electricity consumption at the level of each piece of equipment. However, installing an electricity meter on each piece of equipment in a WWTP is a considerable investment. Indeed, monitoring the individual consumption of a WWTP's equipment requires an average of 20 electricity meters (between 5 and 80 per WWTP), and installing an electricity meter costs an average of 1,000 euros.State of the art Solutions are known for estimating the power consumed by the respective equipment solely from the nominal powers of this equipment which are given by the suppliers, by providing a margin on these powers, just as proposed in the document “Energy consumption of intensive wastewater treatment in France: inventory and variation factors” - Sticker et Al, December 2017, IRSTEA. However, these solutions do not make it possible to identify the drifts in power consumed by the different equipment. One aim of the invention is to limit at least one of the aforementioned drawbacks.Summary of the invention For this purpose, the invention relates to a method for estimating an electrical power consumed by a subset of a group of at least one subset, each subset of the group comprising at least one electrical equipment of a set of electrical equipment, during a period comprising a plurality of elementary periods, the method being implemented by computer and comprising: - estimating the electrical power consumed by the subset using: - individual elementary operating times of each subset of the group during the respective elementary periods and, - total elementary energy consumptions of the set during the respective elementary periods.Advantageously, the method comprises: - determining the individual elementary operating times of each subset of the group during the respective elementary periods, - determining the total elementary energy consumptions of the set during the respective elementary periods. Each of these steps may consist of receiving the total elementary energy consumption values ​​and / or individual elementary operating times and / or calculating the total elementary energy consumptions and / or individual elementary operating times from received values ​​and / or received measurements.Advantageously, the step of estimating the electrical power comprises: - determining, from the total elementary energy consumptions and the individual elementary operating durations, a set of at least one parameter of a parametric model modeling: - a distribution, over the elementary periods, of values ​​of individual elementary powers consumed, during the respective elementary periods, of each subset of the group, - for each elementary period, the total elementary energy consumption during the elementary period, by a sum of a possible constant and the individual elementary energy consumption of each subset of the group during the elementary period.Advantageously, at least when a number of activation / deactivation of at least one subset of the group during all the elementary periods is greater than a main threshold and possibly only when a group of at least one subset of equipment with variable supply frequency of the group (G) is non-empty, the model is a probabilistic graphical model. Advantageously, the parametric model models values ​​of the individual elementary powers consumed, during the respective elementary periods: - of each subset of a first subgroup of the group by a multimodal distribution and / or - of each subset of a second subgroup of the group by a unimodal distribution, - and possibly of each subset of at least one equipment with variable supply frequency of the group (G) by a Gaussian distribution whose mean is a polynomial model of a supply frequency of the subset.Advantageously, the method comprises the following steps: - grouping subsets of the group so that the first subgroup consists of each subset of the group having a number of activations / deactivations during all the elementary periods greater than a predetermined main threshold and the second subgroup consists of each subset having a number of activations / deactivations during all the elementary periods less than or equal to the main threshold. Advantageously, the method comprises the following step: - determining a number of subsets of the first subgroup so that the model is a probabilistic graphical model only when the first subgroup is non-empty.Advantageously, at least when no subset of the group has an activation / deactivation number during all the elementary periods greater than a predetermined main threshold, the parametric model is a linear model. Advantageously, the method comprises the following step: - determining a number of subsets having an activation / deactivation number during the set of elementary periods greater than the main threshold, so that the parametric model is linear only when no subset of the group has an activation / deactivation number during the set of elementary periods greater than the main threshold. Advantageously, the method comprises a step of prior grouping of the equipment of the set according to a correlation criterion between states of the equipment, using state information of the respective equipment, so as to form said group of at least one subset.Advantageously, the method comprises a step of initial estimation of the energy consumption of another set of at least one other equipment from a nominal power of the other equipment and status information of the other equipment. Advantageously, the method comprises an initial verification step implemented for each equipment of a global set, consisting of verifying whether the equipment complies with a set of at least one importance criterion using a predetermined nominal power of the equipment and / or status information of the equipment, the other set of at least one other equipment being made up of each equipment of the set not complying with the set of at least one importance criterion. In one embodiment, the global set comprises each equipment of the set E.Advantageously, the method comprises, prior to the estimation step, a step of identifying each item of equipment in the overall set having a total individual operating time less than or equal to a predetermined maximum portion or greater than or equal to a predetermined minimum portion of the duration T, and a step of removing each item of equipment identified during the identification step from the overall set. The invention also relates to a computer program product stored in a non-transitory computer-readable medium, said computer program product comprising code instructions which, when the program is executed by a computer, cause the computer to execute the method according to the invention. The invention also relates to a data processing device comprising a processor configured to execute a method according to the invention.The invention also relates to an industrial system comprising: - a set of equipment comprising a group of at least one subset, each subset of the group comprising at least one electrical equipment of a set of electrical equipment, - an electric meter, - a control unit for the electrical equipment capable of activating and deactivating each equipment of the set, - a calculation unit configured to implement the method according to the invention, to estimate an electrical power consumed by a subset of the group, the total elementary energy consumptions of the individual elementary operating times of each subset of the group during the elementary periods being derived from energy measurements delivered by the electric meter and the individual elementary operating times of each subset of the group during the respective elementary periods being derived from the control unit.Brief description of the figures Figure 1 schematically represents the equipment of a wastewater treatment plant according to an exemplary embodiment, Figure 2 schematically represents an example of an industrial system according to the invention, Figure 3 represents a flowchart of the steps of a method according to an exemplary embodiment of the invention, Figure 4 schematically represents an example of a Bayesian network modeling an elementary energy consumption and individual elementary powers individually and relative to each other, Figure 5 represents a flowchart of steps of a stochastic approximation algorithm of expectation-maximization according to an exemplary embodiment. Figure 6 represents a flowchart of steps of a method according to another exemplary embodiment.Detailed Description The present invention relates to a computer-implemented method for estimating the electrical consumption of equipment in a global set EG of equipment during a predetermined monitoring period denoted T. Power is understood to mean electrical power in the present patent application. Equipment and industrial system The global set EG of equipment EQ. mwith m integer and m between 1 and M includes, for example, equipment of an industrial system, for example a WWTP or a subset of equipment of a WWTP. A wastewater treatment plant (WTP) comprises several pieces of equipment and functional units allowing the treatment of municipal and / or industrial wastewater by intensifying natural processes. These pieces of equipment vary according to the size of the wastewater treatment plant, the level of treatment required and the specific technologies used. The overall assembly EG performs this function. A WWTP typically includes a pretreatment unit UPT which aims to remove solid debris, coarse matter and undesirable elements from the raw water before proceeding to the main treatment steps.The pretreatment steps and equipment may vary depending on the specific design of the treatment plant, for example: a pump E1, which is supplied via a lifting station to equipment E2, including, for example, a screen to remove different types of debris of different sizes and / or a grit / oil separator to separate sand from floating matter (oils, grease and / or hydrocarbons) and / or a sand washer and / or a biological grease treatment device. A WWTP advantageously, but not necessarily, includes a primary treatment unit UTP. This step typically removes finer suspended matter through flocculation / coagulation / settling and flotation processes - with the injection of chemical reagents depending on the desired degree of treatment, for example using a decanter.A WWTP typically includes a secondary treatment unit UTS used to perform wastewater treatment by biological and / or physicochemical means, exploiting the capacity of certain bacteria or biomass to eliminate pollutants dissolved in the water obtained at the outlet of the primary treatment unit, such as carbon pollution and / or nitrogen compounds and / or phosphorus compounds. The secondary treatment unit UTS includes, for example, at least one source of pressurized air, for example an E3 booster configured to inject air into one of the biological treatment basins of the secondary unit UTS to adapt the oxygen supply to variations in pollution loads, and to the needs of the biomass for the treatment of pollution. The biomasses are thus implemented in one or more biological treatment basins, under different conditions (anoxic, aerobic, anaerobic, etc.) suitable for their metabolism.Biomass develops in the form of free biological sludge (activated sludge process) or in the form of biofilm (biomass supported by biofilm supports or self-supported, i.e. granular sludge). The secondary treatment unit UTS also includes at least one agitator E4 to mix the water and, for example, several brush aerators E5, E6 for the sludge. The tertiary treatment unit UTT implements a step of separating the mixture from the biological and / or physicochemical treatment tank(s). It thus makes it possible to separate sludge on the one hand, and purified water on the other. It includes a separator E7 which can take the form of a flotation system, a filter, a membrane integrated into the biological treatment tank, or a decanter external to the biological treatment tank. The purified water can undergo post-treatment (for example filtration, UV, etc.) or be used directly at the outlet of the separator, or discharged into the natural environment. The WWTP typically includes a sludge treatment unit UB configured to thicken the sludge extracted from the units (possibly UTP), UTS, UTT by increasing their concentration. The sludge thickener can be static (gravity) or dynamic. This thickener can, for example, be in the form of a drainage table, a centrifugation system, a gravity thickener, a drum or even a flotation unit. The thickened sludge from the sludge treatment unit UB can be treated for example using means in the form of a thermal hydrolyser, a digester, dewatering equipment such as centrifuges, belt filters, filter presses, followed or not by chemical (liming) and / or biological (composting) stabilisation and / or thermal drying and / or incineration.The WWTP advantageously comprises an odor treatment unit UTO from the secondary treatment unit UTS and the sludge UB. This odor treatment unit UTO comprises, for example, at least one fan E8 and / or a deodorization tower E9 and / or a reagent station E. 10. The WWTP may also include an air conditioning unit UC to regulate the temperature of the WWTP premises, for example offices. The WWTP includes lighting equipment UE, making it possible to illuminate the WWTP premises. Alternatively, the WWTP includes some of the aforementioned equipment. The WWTP may include other equipment. The invention applies to equipment in other types of plants. In a non-limiting manner, the plant may be an energy recovery unit. An energy recovery unit is a plant supplying a heat network or producing electricity through the incineration of waste. Such a unit includes one or more waste combustion furnaces. Each furnace includes a burner to ensure the temperature of the combustion chamber rises to incineration conditions. The furnace may be gas or electric and, in the latter case, consume electrical energy when the furnace is started.The heat of the gases and fumes resulting from combustion (temperature between 850°C and 1000°C) is advantageously transferred, via at least one transfer device comprising for example a fan, to a heat transfer fluid (generally water) circulated, via a circulation system, in a recovery unit, comprising for example at least one boiler and / or a turbine to recover the heat energy and recover it, for example in the form of electricity or heat. The energy recovery unit advantageously comprises a unit for treating the fumes from the incineration unit to capture the pollutants included in the fumes before the discharge of the treated fumes into the atmosphere. Each of these units advantageously comprises at least one fan and / or at least one pump to circulate a fluid in the energy recovery unit. These devices consume electrical energy.The energy recovery unit advantageously includes a unit for recovering bottom ash, the solid residue from the combustion of waste. This unit typically includes mechanical transfer equipment for this bottom ash, for example conveyor belts which consume electrical energy. Just like WWTPs, an energy recovery unit typically includes an air conditioning unit to regulate the temperature of the premises of the energy recovery unit, for example offices, and a lighting unit, allowing the premises of the recovery unit to be lit. Figure 2 shows an example of a global set EG of EQ equipment. m(m integer with m = 1 to M) of an industrial system SI. This global set EG includes, for example, equipment of a STEP, for example the equipment of the secondary treatment unit UTS, the equipment EQ1, EQ2, EQ3, EQ4 corresponding, for example, respectively to the equipment E3, E4, E5, E6 of figure 1 and the lighting unit UE of figure 1 noted EQ5 in figure 2. In the global set EG of equipment EQ m with m = 1 to M, we consider a set E of equipment EQ i with i = 1 to N, N being less than or equal to M, taken from the EQ equipment m of the global set EG. In the non-limiting example of Figure 2, M is equal to 5 and N is equal to 4. In the method according to the invention, a group G of at least one subset SE is considered j , with j integer and j= 1 to J, J being greater than or equal to 1 and less than or equal to N. Each subset SE j includes at least one piece of equipment from set E. Each EQ piece of equipmenti belongs to only one of the SE subsets j . In other words, the SE subsets j are disjoint. In the non-limiting example of Figure 2, J = 2. Furthermore, in the non-limiting example of Figure 2, the EQ equipment m are the EQ equipment i for m= 1 to 4. The industrial system SI comprises, for example, an electricity meter CE, a control unit UCO and a calculation unit UCA. The electricity meter CE is capable of making measurements ME(k) of an energy representative of the total energy consumption during different elementary periods T k of a period T. The period, noted T, includes K elementary periods noted T k of the same elementary duration ^^^^ ^^^^ , with K an integer greater than 1. Advantageously, the period T is made up of the K elementary periods T k . More generally, the period T includes K elementary periods T k. In a variant, the duration of the period T is greater than the sum of the durations of the elementary periods. For example, at least two elementary periods of the period T are spaced apart in time from each other. K may depend on the period of provision of the information used in the method according to the invention. In a non-limiting example, the period T is 1 month and the elementary duration is 10 minutes. The elementary duration is adapted by a person skilled in the art to the equipment of the overall assembly and in particular to the average frequency of change of state (i.e. activation / deactivation) of this equipment. The control unit UCO is capable of controlling each piece of equipment of the assembly E to make it pass from an activated state (i.e. running) to a deactivated state (i.e. stopped), and vice versa.This UCO control unit is configured to deliver activation / deactivation commands for the respective equipment or, at least, subsets of equipment of the set E so as to individually activate / deactivate the respective equipment. Advantageously, the UCO control unit is able to adjust the operating level of at least one of the equipment of the set E, for example the speed of a motor, or the power level of a booster (maximum power or not). For example, the UCO control unit is able to adjust an operating frequency of at least one of the equipment EQ. i of the set E, namely at least one piece of equipment with variable supply frequency EQ f among the N EQ equipment i of the set E with f integer f = 1 to F with F greater than or equal to 1. Advantageously, the UCA calculation unit has individual elementary operating times a j(k) respective SE equipment subsets j on the respective elementary periods T k . These elementary operating times have j (k) can be calculated by the UCA calculation unit from status information allowing these individual elementary operating times a to be calculated j (k) received by the computing unit UCA. This status information is, for example, transmitted to the computing unit UCA from the control unit UCO. This status information is, for example, activation / deactivation times {tc m} EQ equipment m or the activation / deactivation times {tc j} SE subsets j respective of the set E are transmitted to the calculation unit UCA from the control unit UCO and the calculation unit UCA is configured to calculate the elementary operating times t kjfrom the activation / deactivation times considered. By activation times of a piece of equipment or a subset of equipment, we mean the times when the equipment changes from its inactive state to its active state and by deactivation times of the equipment, we mean the times when the equipment changes from its active state to its inactive state. By activation / deactivation times of a piece of equipment, we mean the activation times and the deactivation times of the equipment. This also applies to subsets of equipment. In another example, the state (active or inactive) of each piece of equipment or each subset of equipment SE j of the set E is transmitted at regular time intervals (shorter than the elementary periods T k to the UCA calculation unit for the entire duration T and constitute the status information. The UCA calculation unit calculates the individual elementary operating durations a j(k) from these states. Alternatively, the UCA calculation unit directly receives the elementary operating times a j (k) SE equipment subassemblies j , on the elementary period T k The UCA calculation unit is configured to implement the method according to the invention from total elementary energy consumptions CET(k) and individual elementary operating durations a j (k) SE subsets j respective. Advantageously, the individual elementary operating times a j (k) come from the control unit UCO. Advantageously, the total elementary energy consumptions CET(k) are the elementary measurements ME(k) or are determined, for example by the calculation unit UCA, from respective elementary measurements ME(k) taken by the electricity meter CE during the respective elementary periods T k. For example, the calculation unit UCA is configured to estimate the total elementary energy consumptions CET(k) from the respective elementary measurements ME(k) delivered by the meter CE. These estimates are then the total elementary energy consumptions CET(k) used in the method according to the invention. Method according to the invention As visible in figure 3, the method according to the invention comprises a step EPIND of estimating at least one power consumed by at least one subset SE j of group G, from: - the individual elementary operating times a j (k) of each subset SE j of group G during elementary periods T k respective of the period T, - and of total elementary energy consumption CET(k) during the elementary periods T krespective. This method does not require the use of a frequency behavior of a voltage measured at the terminals of the set E nor of a current flowing through this set E. In other words, this method uses individual elementary operating times a j (k) and elementary total energy consumptions CET(k), for example elementary energy consumption measurements, without using frequency behavior of a voltage at the terminals of the set of equipment or of a current flowing through this set of equipment to determine individual powers consumed by subsets SE jof equipment making up this set E. An advantage is that it makes it possible to distinguish the power consumed by different subsets of equipment or equipment with similar operating principles (compressor, motor, pump) which give rise to similar frequency signatures. This method is precise, inexpensive, and particularly suitable for sets of equipment such as factory equipment, such as wastewater treatment plants. Indeed, this method only requires the presence of an electric meter CE making it possible to generate measurements ME(k) representative of the total elementary energy consumption CET(k) of the set E during the respective elementary periods T k or allowing the calculation unit UCA to estimate the total elementary energy consumption CET(k) of the set E during the respective elementary periods T kand the arrangement, by the UCA calculation unit of the individual elementary operating times a j (k) SE subsets j of equipment. . The invention takes advantage of the activation / deactivation times of the various equipment in the set, allowing the calculation of the individual elementary operating times a j(k) conventionally available in a factory in which the equipment is controlled automatically, to estimate the elementary powers of the equipment or subsets of equipment. Another advantage of the method is that the EPIND estimation step of the method can be implemented without knowing the nominal power of each subset. In other words, in a particular embodiment, the EPIND estimation step advantageously does not use nominal powers of equipment or subsets of equipment of the set E. This invention also makes it possible to detect a degradation in the energy efficiency of the subsets of equipment. In the remainder of the description, it is considered that the set E comprises several subsets, each comprising at least one piece of equipment, which will be called subsets of equipment even if each of them can comprise a single piece of equipment.The invention also applies when the set E comprises a single subset and / or at least one subset comprises a single item of equipment. The method advantageously, but not necessarily, comprises a step of calculating at least one individual elementary energy consumption of at least one subset SE. j during an elementary period T k , from the individual power consumed during the elementary period and the individual elementary operating duration a j (k) of the SE subset j during an elementary period T k . Advantageously, this step includes, for each subset SE j , the calculation of the individual elementary energy consumption of this subset during the elementary periods T krespective. Disaggregation of the energy consumed Advantageously, during the EPIND estimation step, a parametric model is used modeling the total elementary energy consumption CET(k) during each elementary period T k in a linear combination of the elementary energy consumptions of the SE equipment subsets j respective during the elementary period T k considered. For example, the CET(k) value of the total elementary energy consumption during each elementary period T k is modeled by a sum of a possible constant CST and individual elementary energy consumption CE j (k) respective subsets SE j , during the elementary period T k The model models the individual elementary energy consumption CE j (k) of a subset SE jby a product between an elementary power consumed by the subset SE j during period T k and the individual elementary operating time has j (k) of the subset during the elementary period T k In other words, the model models the individual elementary energy consumption CE j (k) as follows: ^^^^ ^^^^ ^^^^ ( ^^^^) = ^^^ ^ ^ ^^^ ( ^^^^) ∗ ^^^^ ^^^^ ( ^^^^) By individual elementary power P j (k) consumed by a subset of SE equipment j , we mean the electrical power consumed by the SE equipment subset j during the elementary period duration T k . This is, for example, the electrical power absorbed by this subset during this elementary period T k This individual elementary power consumed is considered fixed during the elementary period Tk. The constant CST is, for example, the energy generated by equipment whose state remains substantially the same over the entire duration of the period T. In order to distribute the total elementary energy consumption CET(k) over the different sub-assemblies SE j we advantageously make, during the EPIND step, an a priori on the distributions of the individual elementary powers consumed P j (k), on the elementary periods T k , by the SE subsets j . In other words, the parametric model models the distributions of the individual elementary powers consumed P j (k), on the elementary periods T k , by the respective SE subsets j . The determination of the elementary powers consumed P j (k) by SE equipment subassemblies jrespective therefore includes the determination of the parameters of the parametric model which optimizes, according to a predetermined criterion, the distribution of the total elementary energy consumption CET(k) during the elementary periods T k In summary, we distribute each total elementary energy consumption CET(k) between the different subsets SE j using the elementary operating times a j (k) of the different SE subsets j and by determining the parameters of a parametric model modeling: - the distributions of the individual elementary powers P j (k) SE subsets j by at least one parameter of the model, - and each total elementary energy consumption CET(k) by a sum of a possible constant CST and the individual elementary energy consumptions CE j (k) respective subsets SE j , during the elementary period T k. This step will be described in more detail later in the text. Modeling of total energy consumption The model models the total elementary energy consumption ^^^^ ^^^^ ^^^^( ^^^^) during each elementary period T k by a sum of a possible constant CST and individual elementary energy consumption CE j (k) respective subsets SE j , during the elementary period T k . In other words, the total elementary energy consumption ^^^^ ^^^^ ^^^^( ^^^^) is defined as follows: The value of the total elementary energy consumption ^^^^ ^^^^ ^^^^ ( ^^^^ ) during the elementary period T kis an input data. It is, for example, measured by the CE electricity meter or estimated based on a measurement from the CE meter as explained previously. Initial verification A non-limiting example of implementation of the method according to the invention is shown in Figure 3. The method advantageously comprises an initial verification step VERI, implemented for each EQ equipment m of the global set EG for m integer ranging from 1 to M, consisting of verifying compliance with a set of at least one importance criterion in terms of power and / or number of activations / deactivations and, if at least one criterion of the whole equipment is not respected by said equipment EQ m , an initial estimation step EST1 of CE energy consumption m EQ equipment m . The initial VERI verification of compliance with all of at least one important criterion uses a nominal power PN mpredetermined EQ equipment m and / or EQ equipment status information m during period T or elementary periods T k . The nominal power PN m is typically provided by the manufacturer or accessible in the supplier documentation. In the non-limiting example of Figure 2, the initial VERI verification includes a first VERIa verification of compliance with a first criterion of importance in terms of power from the nominal power PN m EQ equipment m . The criterion of importance in terms of power, noted C1a in the present application, is, for example, respected only when the nominal power PN m EQ equipment mis greater than or equal to a predetermined power threshold SP. Advantageously, the predetermined power threshold, noted SP in this patent application, is between 1% and 2% of the average nominal power of the EQ equipment mof the overall set EG. This makes it possible to exclude from the set E and therefore from precise power calculations, equipment which has little impact on energy consumption drifts due to the low nominal power. For example, in the example of Figure 2, the lighting unit is not very powerful compared to the other equipment in the overall set EG and therefore does not meet this criterion C1a. In a non-limiting example, the threshold is set at 2 kW. In the non-limiting example of Figure 2, the initial verification VERI includes, only when the first importance criterion is met, a second verification VERIb of compliance with an importance criterion in terms of the number of state changes, based on state information, for example the activation / deactivation times of the equipment EQ m during period T or during elementary periods T k. The significance criterion in terms of the number of activation / deactivation of states C1b is, for example, only met when a number of state changes, i.e. the number of activation / deactivation, of the EQ equipment m during period T or during all periods T k , is greater than or equal to a predetermined initial threshold SI. The number of activations / deactivations of a device during a period corresponds to the number of changes of state between the activated state and the deactivated state during the entire duration of the same period. It is the sum of the number of activations and the number of deactivations obtained during the entire duration of the same period. The number of activations / deactivations of a device during all the elementary periods T k is the sum of the activation / deactivation numbers corresponding to the different elementary periods T k. Advantageously, the initial threshold SI is between 5 and 10. This makes it possible to avoid integrating into the set E equipment for which the number of state changes is not sufficient to guarantee good performance of the statistical calculations. Alternatively, the initial verification VERI includes only one of the two verifications VERIa or VERIb. Alternatively, the order of the two verification steps is reversed. It is of course possible to envisage, as a variant or in addition, other criteria depending on the information available. Alternatively, the method does not include the initial verification VERI. Initial estimation The initial estimation EST1 is implemented for each equipment EQ mnot respecting the set of at least one importance criterion. These equipments form a so-called remaining set and noted ER of residual equipment MR, MR being an integer between 0 and M-1. In the example of figure 2, the residual set ER is made up of the lighting unit UE. This step is therefore, for example, implemented only when the remaining set ER is not empty. The initial estimate EST1 of the energy consumption CE m of an EQ equipment m uses the nominal power PN m EQ equipment m and a total operating time T m EQ equipment m during which the equipment considered is in the active state during all the elementary periods during all the elementary periods T k . The total operating time is, for example, calculated from the times the EQ equipment is switched on / off. mduring the period T or of elementary operating times of the equipment during the different elementary periods T k . The total operating time T m of an EQ equipment m is, for example, given by the following formula: where ^^^^ ^^^^ ( ^^^^) is the elementary operating time during which the EQ equipment m is in the active state, during the k ieme elementary period T k , and ^^^^ ^^^^ is the elementary duration of each elementary period T k . The initial estimate EST1 therefore includes the CD calculation of the total individual operating time ^^^^ ^^^^ EQ equipment m from status information, e.g. activation / deactivation times, relating to the EQ equipment m The initial estimate EST1 then includes a calculation CP1 of a product of a predetermined correction factor FC, of ​​the nominal power PNm EQ equipment m and the operating time T m EQ equipment m during the period T or the set of elementary periods T k . In other words, the energy consumption of the CE equipment m during the period T or during all the elementary periods T k is, for example, estimated during step EST1 using the following formula: ^^^^ ^^^^ ^^^^ = ^^^^ ^^^^ ∗ ^^^^ ^^^^ ^^^^ ∗ ^^^^ ^^^^ The correction factor FC is a constant between 0.5 and 1 corresponding to a factor between an actually consumed power and the nominal power. The correction factor is, for example, equal to 0.85. Alternatively and / or in addition, the estimation step includes the estimation of the individual elementary energy consumption CE m (k) EQ equipment m for at least one elementary period T k. We use the following formula advantageously: ^^^^ ^^^^ ^^^^ ( ^^^^) = ^^^^ ^^^^ ∗ ^^^^ ^^^^ ^^^^ ∗ ^^^^ ^^^^( ^^^^) The initial estimate EST1 has the advantage of being inexpensive in terms of calculation and fast. In addition, this initial estimate makes it possible to take into account the consumption of equipment that consumes low power compared to other equipment in the overall set and whose estimate of power consumed by a method of solving a linear or probabilistic regression model is not very precise. Indeed, when the variation in the power generated by a piece of equipment is lower than the random variation in the value generated by another piece of equipment, these models are not able to differentiate these two variations. Advantageously, the initial estimate EST1 uses the nominal power of each piece of equipment.Advantageously, the method then comprises a step EST_CET of estimating the total elementary energy consumption CET(k) by the set E from the measurements ME(k) and the individual elementary energy consumption ^^^^ ^^^^. ^^^^ (^^^^) EQ equipment m of the remaining set ER. The total elementary energy consumption CET(k) by the set E during the elementary period T k is then the energy consumption measured during this elementary period T k from which we subtract the sum of the elementary energy consumption of the equipment in the remaining ER set, i.e.: Package E The EQ Equipment Package E iwith i = 1 to N, N being less than or equal to M, advantageously comprises, when the first verification step VERI is implemented, only on the equipment of the global set EG respecting the set of at least one importance criterion in terms of power and / or number of activations / deactivations C1a, C1b. Thus, the set E comprises only the equipment having a significant power in view of the set of equipment and / or which has state change characteristics, and more precisely the dynamics, necessary for a good estimation of their power from their state change information. In the non-limiting example of Figure 2, the set E does not include the lighting unit UE. Alternatively, the set E comprises all the equipment EQ m, with m = 1 to M. In other words, the set E is the set EG. For example, the method is devoid of the initial verification step VERI. Preliminary grouping REG1 Advantageously, the method comprises, as shown in the non-limiting example of Figure 3, a preliminary grouping REG1 of the equipment EQ i of the set E according to a predetermined prior grouping criterion so as to form the group G composed of at least one subset SE j with j being an integer ranging from 1 to J, J being less than or equal to N. Each subset SE j Group G includes at least one EQ device i of the set E. This preliminary grouping REG1 is carried out from the status information of the EQ equipment i of the set E. The prior grouping criterion is advantageously a correlation criterion between the states of the EQ equipment i of the set E for the entire duration of the set of elementary periods Tk or of the period T. We calculate, for example, for each pair of equipment in the set E, a correlation coefficient ^^^^( ^^^^, ^^^^) given by: where X and Y are the lists of states ^^^^ ^^^^ , and respectively ^^^^ ^^^^ , of a first equipment and respectively of a second equipment during the period T or the set of elementary periods T k . These n states ^^^^ ^^^^ and ^^^^ ^^^^ of index t, with t being an integer ranging from 1 to n, are, for example, taken at regular time intervals. The time interval is less than the duration of the elementary periods. Each state ^^^^ ^^^^ or ^^^^ ^^^^takes the value 0 when the equipment is inactive and the value 1 when the equipment is active. An advantage is to allow the estimation of an individual elementary power consumed and therefore the identification of power drifts of a set of equipment operating almost simultaneously and therefore difficult to distinguish as we will see later. In the non-limiting example of figure 2, at the end of the preliminary grouping step REG1 applied to the global set E composed of the equipment EQ1 to EQ4, we obtain two subsets SE1 and SE2 composed respectively of the equipment EQ1 to EQ3 and the equipment EQ4 corresponding for example respectively to the equipment E3 to E5 and E6 of a secondary processing unit UTS. Status information of a subset SE According to an exemplary embodiment, the status information of a subset SE jare the status information of a piece of equipment in the subassembly. For example, the number of status changes of a subassembly SE j during the duration T or during an elementary period T k is the number of state changes of a piece of equipment in the subset during the period considered. Alternatively, the state information of a subset SE j are obtained from status information of several equipments of the SE subset j , for example of all equipment in the SE subset j . For example, the averages of the equipment status information of the SE subset j The same applies to each elementary operating duration of an SE subset. j during an elementary period T k may be, in a non-limiting manner, an average of the operating times of the equipment of the sub-assembly during the same elementary period T kor the basic operating time of one of the equipments of the SE sub-assembly j during the elementary period. The total individual operating time T j of an EQ equipment j is, for example, given by the following formula: where ^^^^ ^^^^ ( ^^^^) is the elementary operating time during which the SE equipment subset j is in the active state, during the k ieme elementary period T k The same applies to the number of activations / deactivations of each SE subset. jduring a given period which may be, in a non-limiting manner, an average of the activation / deactivation numbers of the equipment of the subsets of equipment which compose it during this period or a number of activation / deactivation of one of the equipment which it comprises during this period. The method advantageously comprises a step of determining a number of activation / deactivation, during the period T or during all the periods T k , of at least one piece of equipment from the status information of this equipment. The method advantageously comprises a step of determining a number of activations / deactivations, during the period T or during all the periods T k, of at least one subset of equipment from the state information of at least one of its equipment. Distribution of individual elementary powers The parametric model comprises a set ^^^^ of parameters comprising a first set of parameters modeling the respective distributions of the values ​​of the individual elementary powers P j (k) SE subsets jrespective. The EPIND estimation step includes determining the parameters of the parametric model. The probabilistic graphical model has the advantage of not requiring, in order to obtain precise estimates of the different individual powers, a strict a priori on the distribution of the individual elementary power values ​​in the sense that it allows the use of general distributions such as unimodal or multimodal distributions. In particular, the probabilistic model does not require considering that the individual elementary powers of each of the equipment are fixed, that is to say, are the same over the entire period T. In other words, this model takes into account the potential variations in the powers consumed by the subassemblies during their operation during the period T, which allows for a precise estimation of these powers.It therefore makes it possible to detect drifts occurring over shorter durations than over a period T. This makes it possible to consider long durations T. This type of solution is suitable for different industrial installations with equipment consuming different powers or for the same industrial installation at different periods between which the powers consumed by the different equipment vary. These consumed powers can vary with different parameters, for example temperature, or over time during the lifetime of the equipment. The powers of the SE subassemblies. j are hidden variables (i.e., variables unknown to the probabilistic graphical model). The model is also defined by the fact that when the subset SE j is deactivated for the entire duration of the period T or of all the elementary periods T k , the individual elementary power P j(k) of the SE subset j corresponding is equal to zero. By a subset SE j deactivated for the entire duration of the period T, we mean a subset SE j whose total individual operating time D j is zero. When the subset SE j is activated for at least part of the period T or of the set of elementary periods, that is to say when its operating time is non-zero over this period or set of periods, the distribution of the values ​​of individual elementary powers P j (k) of this subset SE j is modeled by a parametric graphical model whose parameter values ​​are to be determined. This model is a probabilistic graphical model. The distributions of the values ​​of the individual elementary powers ^^^ ^ ^ ^^^ ( ^^^^ ) SE subsets jrespective can be multimodal or unimodal. The unimodal distribution is defined by a set of so-called unimodal parameters. An example of a unimodal distribution of the values ​​of the individual powers of a subset of SE equipment j is a Gaussian or normal law ^^^^( ^^^^ ^^^^ , ^^^^ ^^^^ ). where the unimodal parameters are ^^^^ ^^^^ which is an average of the normal law and ^^^^ ^^^^ which is a standard deviation of the normal distribution. Another example of a unimodal distribution of the values ​​of the individual powers of a subset of SE equipment j is a gamma distribution or gamma law defined by the following unimodal parameters ^^^^ ^^^^ is the shape parameter of the gamma distribution associated with the subset SE j . ^^^^ ^^^^ is the intensity parameter of the gamma distribution associated with the SE subset jAn example of a multimodal distribution defined by so-called multimodal parameters is a mixture of gamma laws ^ ^^^^^^^^^, ^^^^^^^^( ^^^^^^^^, ^^^^, ^^^^^^^, ^^^^) where ^^^^ is an integer and L is an integer greater than 1 and corresponds to the number of modes of the mixture of gamma distributions. In other words, in general, the multimodal distribution is a mixture of distributions of different modes where ^^^^ is an integer with ^^^^ ranging from 1 to L. The mixture of gamma distributions is, for example, defined by a set of the following so-called multimodal parameters: ^^^^^^^^, ^^^^, ^^^^^^^^, ^^^^, ^^^^^^^, ^^^^. ^^^^^^^^, ^^^^is the shape parameter of the distribution of the mode ^^^^, that is to say of the mode of order ^^^^, of the distribution associated with the subset SE j . ^^^^^^^^, ^^^^is the intensity parameter of the distribution of the order mode ^^^^ of the distribution associated with the subset SE j. ^^^^^^^^, ^^^^is the weight parameter of the order mode ^^^^ in the distribution associated with the subset SE j . Distribution of total elementary energies Advantageously, the set ^^^^ of parameters also includes a second set of at least one parameter defining a distribution of the total elementary energy consumptions CET(k) for k = 1 to K by one or more parameters and at least one parameter of this distribution is advantageously defined during the step of determining the parameters of the model. This distribution is, for example, a Gaussian defined by the following two parameters: a standard deviation σ and an average of the elementary energy consumption ^ � ^^^( ^^^^) .The values ​​of the total elementary energy consumption CET(k) are therefore defined by a normal law ^^^^ depending on these two parameters. A parameter modeling the distribution of the values ​​of the total elementary energy consumption CET(k) can be predetermined, for example ^ � ^^^( ^^^^) , while the other parameter(s) are estimated when estimating the model parameters or all parameters can be estimated when estimating the model parameters. Distribution types other than Gaussians can of course be considered. One can, for example, define a gamma distribution or a mixture of gammas. Probabilistic graphical model At least when an activation / deactivation number of at least one SE subset j during all elementary periods T k is greater than a predetermined main threshold, the parametric model is a probabilistic graphical model. It should be noted that when the period T is made up of the elementary periods T k, the number of activation / deactivation of the SE subset j during all elementary periods T k is the number of activations / deactivations during the period T which is not the case when the period T comprises a set of elementary periods whose duration is less than the total duration (for example when they are disjoint or when they do not cover the entire duration T). In the embodiment of Figure 3, the parametric model is a probabilistic graphical model only when a number of activations / deactivations of at least one subset SE j during all elementary periods T kis greater than a main threshold. Advantageously, the main threshold is greater than the initial threshold as we will see later. Alternatively, the main threshold is the initial threshold. Alternatively, the main threshold is zero. In other words, we use a probabilistic graphical model, regardless of the activation / deactivation numbers of the different SE subsets j . The probabilistic graphical model is, for example, a Bayesian network as represented in Figure 4, in which the hidden variables ^^^ ^ ^ ^^^ ( ^^^^ ) , the set of parameters, as well as the individual elementary operating times ^^^^ ^^^^( ^^^^) and the total elementary energy consumption are the nodes NO of the network. These nodes NO are connected two by two by arcs A as represented in figure 4 so as to define the parametric model described previously. Advantageously, the estimation step EPIND includes a step denoted AL of implementation of a stochastic approximation of expectation maximization algorithm also called SAEM, acronym for the English expression “Stochastic Approximation of Expectation-Maximization”. This type of algorithm is known to those skilled in the art and will be described succinctly in the patent application. The implementation of the SAEM algorithm includes the following step: - initialize IN the set ^^^^ of parameters, that is to say the values ​​of the parameters of the model and the individual elementary powers consumed ^^^ ^^^^^ ( ^^^^ ) for k= 1 to K and for j= 1 to J by the subsets of equipment SE j, that is, the values ​​of the hidden variables. The value of each individual elementary power ^^^ ^ ^ ^^^ ( ^^^^ ) for k= 1 to K of a given subset SE j is, for example, initialized to a sum of a constant and the nominal powers of the equipment constituting the SE subset j . The constant may be zero. Alternatively, the values ​​of the individual elementary powers are initialized to values ​​within a predetermined power range, for example to the same power value. This initialization is, for example, carried out randomly. The following parameters may, in a non-limiting manner, be initialized in the following manner, during the initialization step IN. For each subset SE j whose distribution of individual elementary power values ​​is modeled by a gamma mixture defined by the following parameters: ^^^^^^^^, ^^^^, ^^^^^^^, ^^^^, we initialize these parameters using the following: where ^^^^^^^^, ^^^^is the expectation of the distribution of the mode ^^^^ of the subset SE j and whose value is initialized to the nominal power of the SE subset j by multiplying by a variable with a value within a predetermined interval. The predetermined interval extends, for example, from 0.7 to 1.3. The value is, for example, randomly set within the interval. ^^^^^^^^, ^^^^is the standard deviation of the distribution of the mode ^^^^ of the subset SE j and is initialized to a constant, for example 1. ^^^^^^^^, ^^^^= 1 / Nmode where Nmode is the number of modes of the gamma mixture. ^^^^^^^^, ^^^^and ^^^^^^^^, ^^^^are calculated using the above equations and the initial values For each SE subset jwhose distribution of individual elementary power values ​​is modeled by a Gaussian distribution ^^^^ ^^^^ is initialized to a value ^^^^ ∗ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ where ^^^^ is a predetermined constant, for example 0.1 and where ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ is the nominal power of the SE subassembly j . The rated power of SE j is either known or calculated from the nominal powers of the equipment constituting SE j . This is, for example, the sum of these nominal powers. For the distribution of the values ​​of total elementary energy consumption: ^^^^ 2 is, for example, initialized to 1. The value of ^ � ^^^( ^^^^) is, for example, initialized to the sum of the individual elementary consumptions estimated by the SE subsets j during the elementary period T k, calculated from the initial values ​​of these individual elementary powers ^^^ ^ ^ ^^^ ( ^^^^). The constant CST can be initialized to a predetermined percentile of an elementary energy consumption CET(k). The constant CST is, for example, set to the 5th percentile of the energy consumption CET(k). The step AL of implementing the stochastic approximation algorithm of expectation maximization includes the following set of steps EE repeated, as long as a stopping criterion is not respected by the current values ​​of the parameters: - generate GEN a set P of values ​​of the hidden variables distributed, over the set of elementary periods T k, according to the distributions defined by the current values ​​of the parameters, - calculate CAE the expected likelihood of the model defined by the current values ​​of the set ^^^^ of parameters, from these samples, - implement EMA a maximum likelihood estimator of the parameters so as to update the current values ​​of the parameters. During the first implementation of the set of steps EE, the current values ​​of the parameters are the initial values ​​of the parameters defined during the initialization step. The set P generated during the generation step GEN includes for each subset SE j , a set of K values ​​of individual elementary powers ^^^ ^ ^ ^^^ ( ^^^^ ) distributed according to the distribution defined by the current values ​​of the CET(k) parameters, and ^^^ ^ ^ ^^^′ ( ^^^^ ) for everything ^^^^′ ^^^^ and being an integer between 1 and J. This step is, for example, implemented by a Markov chain Monte Carlo algorithm also called MCMC (acronym for the English expression “Markov chain Monte Carlo”), for example a Hamiltonian Monte Carlo algorithm, also known as Hybrid Monte Carlo, to generate samples of the distributed hidden variables, over the set of elementary periods T k , according to the distributions defined by the values ​​of the current parameters. In an exemplary embodiment, during the GEN step, an MCMC algorithm is implemented Q times (where Q is an integer greater than zero) so as to generate Q sets P of values ​​of the hidden variables and the set P generated during the GEN step is an average of the Q sets generated. The likelihood of the model ^^^^( ^^^^) defined by the set of ^^^^ is given by the following formula: where J is the number of SE equipment subsets j. K is the number of elementary periods. ^^^^ ^^^^ is the set of distribution parameters that model the distribution of individual elementary powers of SE j . ^^^^( ^^^^ ^^^^ ) is a priori probability of the set of parameters ^^^^ ^^^^ . ^^^^( ^^^ ^ ^ ^^^ ( ^^^^) | ^^^^ ^^^^ )) is a posteriori probability of a set of powers knowing the parameters ^^^^ ^^^^ distributions of the SE subset j . ^^^^( ^^^^^^^^ ^^^^ ^^^^ ^^^^( ^^^^) / | ^^^^1( ^^^^), ... , ^^^^ ^^^^ ( ^^^^)) is an a posteriori probability of an elementary power consumed during the elementary period T k knowing the elemental powers ^^^ ^ ^ ^^^ ( ^^^^) of the different equipment during the elementary period T k . The expected likelihood ^^^^(( ^^^^( ^^^^)) is, for example, calculated using the following formula: where Q is the number of implementations of the MCMC algorithm, ^^^^( ^^^^) ^^^^is the likelihood calculated during the qth implementation of the MCMC algorithm to generate the powers with q integer ranging from 1 to Q. At the end of the EMA step, the current values ​​of the parameters are those which maximize the likelihood, according to the implemented estimator. In other words, the current values ​​of the parameters are those which make it possible to obtain the highest value of the likelihood by implementing the estimator. It should be noted that the implementation of the AL algorithm is both a step of determining the values ​​of the model parameters, that is to say a step of parameterization of the model, and the values ​​of the hidden variables. These values ​​of the model parameters are the last current values ​​of the parameters generated during the expectation maximization algorithm. In other words, these are the values ​​of the parameters obtained during the last implementation of the EMA step. The values ​​of the individual elementary powers ^^^ ^ ^^^^( ^^^^) are also the last current values ​​of these powers, i.e. the values ​​obtained during the last implementation of the GEN step. The stopping criterion is, for example, respected when the set EE of steps has been implemented a number of times equal to a predetermined maximum number, this number being an integer greater than 1. In a variant, the stopping criterion is a criterion of difference between the two likelihoods calculated during the last implementation of the set of steps and during the previous implementation, less than or equal to a predetermined threshold. In a variant, the stopping criterion is a criterion of difference between the two likelihoods calculated during the last implementation of the set of steps and during the previous implementation less than or equal to a predetermined threshold, for a minimum number of successive implementations of the set of steps. The minimum number is advantageously greater than or equal to 2.In one embodiment, step EPIND comprises, at the end of step AL, the implementation of a step ESTE0 for estimating the elementary energies E. j consumed by SE subsets j respective from the individual elementary powers ^^^ ^ ^ ^^^ ( ^^^^) and individual elementary total operating times ^^^^ ^^^^ ( ^^^^) SE subsets j during period T. This step includes, for each subset SE j , the realization of the sum of the products, carried out for each elementary period of its individual duration of operation ^^^^ ^^^^ ( ^^^^ ) by its individual elemental power ^^^ ^ ^ ^^^ ( ^^^^). REG2 Grouping In a particular embodiment, the EPIND step comprises a secondary grouping REG2 of the SE subsets jof group G: - into a first subgroup G1 made up of each subset having an activation / deactivation number, during all the elementary periods T k , greater than a main threshold, - and in a second subgroup G2 consisting of each subset having an activation / deactivation number, during all the elementary periods T k , is less than or equal to the main threshold. This step REG2 is carried out using the activation / deactivation numbers, during all the elementary periods T k , SE subsets j respective of the group G. Each of these groups G1, G2 can be empty or made up of all the subsets SE jof group G. Advantageously, the distribution of the values ​​of the individual elementary powers consumed by each subset of the first group G1 is modeled by a multimodal distribution and the distribution of the values ​​of the power of each subset of the second group G2 is modeled by a unimodal distribution. These distributions are, for example, used in the model and therefore during the implementation of the expectation maximization algorithm to determine the values ​​of the parameters and hidden variables of the model. Alternatively, the method is devoid of secondary grouping. In the latter case, the distributions, used in the model, of the values ​​of the individual elementary powers of all the equipment can be unimodal or multimodal. Linear regression model REGL In the non-limiting embodiment of Figure 2, the method comprises a step DNB of determining the number of subsets SE jpresenting a number of activation / deactivation during all the elementary periods T k , greater than the main threshold. For example, we determine the number of subsets included in the first group G1. This step also includes a step of comparing the number of activations / deactivations during all the elementary periods T k greater than the main threshold, i.e. when the first group G1 is empty. When no subset SE j of group G does not present a number of activation / deactivation during all the elementary periods T kgreater than the main threshold, i.e. when the first group G1 is empty, the parametric model is a linear regression model. The EPIND estimation step includes the REGL implementation of a parameter estimation algorithm of the linear regression model. The use of the linear model makes it possible to estimate the individual elementary powers even in the case where the number of information (number of activation / deactivation) of the respective subsets is limited. Furthermore, using this linear model only when no subset meets the criterion of the number of state changes has the advantage of not degrading the estimation of the powers of the multimodal subsets.Furthermore, the parameter estimation algorithms of a linear regression model are relatively easy to implement, which limits the cost in terms of calculations and makes it possible to obtain substantially the same results as using an algorithm to determine the parameters of a probabilistic graphic model when the equipment is unimodal. In this type of parametric model, it is assumed that the individual elementary powers are considered fixed over the duration T, that is to say that the individual elementary powers of each subset are the same during all the elementary periods, that is to say whatever K. In other words, each subset SE. j has an individual power P jfixed over all elementary periods or for the entire duration T. In an exemplary embodiment, the linear regression model models the total energy consumed during the elementary period T k by a sum of a constant and the consumed: where CST is a constant, and ^^^^ ^^^^ ( ^^^^ ) is the operating time of the SE subset j during the elementary period T k. The parameters of the linear model are ^^^ ^ ^ ^^^ and ^^^^ ^^^^ ^^^^. The algorithm for estimating the parameters of the linear regression model is, for example, an algorithm for solving the linear regression model by a least squares method or a least squares algorithm. This algorithm is an optimization algorithm that calculates, for the subsets SE j an estimate of their individual powers P j , which minimize the following quantity ^^^^ ^^^^: The method is, for example, a non-negative least squares method, also called "Non-negative least squares" or NNLS in Anglo-Saxon terminology. In this method, we constrain the value of the individual power P j to be a positive value. Alternatively, the value of the individual power P j is constrained to be in a predetermined interval: ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ≤ ^^^ ^ ^ ^^^ ≤ ^^^^2 ∗ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ Advantageously, the least squares method is under-constrained. This method is also called "constraint least squares" in Anglo-Saxon terminology. This makes it possible to limit the space of solutions. ^^^^ ^^^^ ^^^^ ^^^^ ^^^^is a predetermined nominal power of the subset SE. n1 and n2 are positive predetermined parameters such that n2 is greater than n1. n1 is positive and less than or equal to 1 and n2 is positive and greater than 1. The values ​​are set according to the capacity of each subset to operate at a power far from its nominal power. For example, n1 is equal to 0.5 and n2 is equal to 2. The EPIND step advantageously, but not necessarily, includes a step of calculating ESTE the elementary energies E j consumed by SE subsets j respective from the individual powers P j and total individual operating times D j SE subsets j during period T. The elementary energy E j consumed by each SE subset j , during period T is given by the following formula: ^^^^ ^^^^ ^^^^ = ^^^ ^ ^ ^^^ ∗ ^^^^ ^^^^^ Alternatively, the method lacks the DNB verification step and comprises the REGL implementation of the linear regression algorithm to estimate the individual powers of the SE subsets jof group G and possibly the ESTE step described above. However, the previously described embodiment is more advantageous or a disadvantage of the linear model lies in the assumption of the unimodality of the equipment subsets for equipment which may have several operating modes or whose power is intended to vary continuously over a range of values. Variant As a variant, the method comprises a step of determining the parameters of a linear regression model modeling, for each elementary period, the total elementary energy, by a sum of a constant CST and the individual elementary energies consumed by the equipment subsets of the second group. Each individual elementary energy corresponds to a product of a fixed individual power and an elementary duration of operation of the equipment over the elementary period.The EPIND estimate includes an estimate of the total energy consumed by the subsets of the second group G2 from the elementary energies E. j of the subsets of the second group G2. The EPIND estimation then includes a step of correcting each value of CET(k) by subtracting the estimate of the total energy consumed by the subsets of the second group G2 from the corresponding measure of energy consumption. The EPIND estimation then includes a step of determining a set of parameters of a probabilistic graphical model modeling the distributions of the values ​​of the corrected total elementary energy consumption CET(k) and the distributions of the values ​​of the individual elementary powers of the subsets SE j respective of the first group G1 during the respective elementary periods T kby multimodal distributions. The method according to the invention is independent of the values ​​of energy consumption and operating times of the equipment during at least one period prior to period T. Only measurements of energy and activation / deactivation times of the equipment obtained during period T are used. Such a method is more generic, that is to say, applicable to sets E or global sets EG of equipment that are very different from each other. Detection of devices being almost always in the same state Advantageously, the method comprises a step of identifying the equipment, which, in the global set EG, have a substantially identical state (activated or deactivated) throughout the period T.For this purpose, for example, each piece of equipment is identified having a total individual operating time (i.e., a total operating time during the period T) less than or equal to a predetermined maximum portion or greater than or equal to a predetermined minimum portion of the duration T. The maximum portion is, for example, between 0.1% and 1% and the minimum portion is, for example, between 99% and 99.9%. The method also advantageously comprises a step of removing each piece of equipment identified from the global set EG. These steps are advantageously implemented before the estimation step EPIND. This makes it possible to remove, from this step, the equipment likely to disturb the model. The energy consumed by this equipment is found in the constant CST. These steps are, for example, implemented in addition to or instead of the steps VERI, EST1 and EST_CET.Variable power frequency equipment According to one embodiment, the overall assembly EG comprises at least one EQ equipment. h with h= 1 to H, H being less than or equal to N with variable supply frequency having a motor whose supply frequency is controllable so that the rotational speed of the motor and therefore the power consumed by the equipment are controlled when the equipment is in the activated state. Each EQ equipment h with h= 1 to H is one of the EQ equipments iof the overall assembly EG. For example, when the equipment is in the activated state, the supply frequency of the motor is adjustable so that the speed of the motor of the equipment and therefore the power consumed by the equipment are likely to vary continuously over a continuous range of values ​​between 0 and a maximum value. In order to adjust the power consumed by the equipment with variable supply frequency, the control unit UCO comprises one or more frequency converters that can be remote or integrated at least partially into the equipment with variable supply frequency. The control unit UCO is, for example, configured to adjust the power consumed by each equipment with variable supply frequency by controlling the frequency converter associated with this equipment so as to adjust the supply frequency of the motor of this equipment, which makes it possible to adjust the power consumed by this equipment.At least one variable power supply frequency equipment is, for example, taken from a motor, a pump, a booster, a compressor. Thus, the power consumed by adjustable power equipment is likely to vary during the period T even if the equipment remains, throughout the period T, in the activated state. Thus, this type of variation in consumed power cannot be estimated precisely, for example by means of a multimodal distribution model, using only the elementary operating times of the equipment. This estimation is not possible when removing the equipment that remains in the same state throughout the duration T. According to an exemplary embodiment shown in FIG. 6, the estimation step EPIND2 comprises a step RET of removing each variable power supply frequency equipment from the overall set EG so as to form a set EV of H variable power supply frequency equipment EQ. h. The removal step is, for example, carried out from a list of power variation equipment stored in a memory of the calculation unit UCA or more generally of the system. For this purpose, the estimation step EPIND2 comprises, for example, prior to the initial verification step VERI, a step IDE of identification of the equipment with variable power supply frequency. The global set EG updated by the removal step RET comprises only equipment with fixed power supply frequency or for which the variable nature of the power supply frequency is not known. The initial verification step VERI, when it exists, is then carried out only on equipment with fixed power supply frequency. The step REG1 is then, for example, carried out on a set E comprising only the equipment of the global set EG with fixed power supply frequency respecting the at least one importance criterion.At the end of the grouping step REG1, we obtain a group G of subset SEF. f with f= 1 to F of equipment with fixed supply frequency. Alternatively, the removal step RET is implemented after the possible step VERIa and before the possible step VERIb. In the rest of the text, by frequency of an equipment, we mean the supply frequency of the motor of the equipment. The estimation step EPIND2 further differs from the estimation step EPIND in that it includes, for example, but not necessarily a grouping step REG3 of equipment with variable supply frequency EQ h of the EV assembly, into subassemblies called “variable supply frequency” from the frequency values ​​of this equipment so as to form a group, called variable supply frequency GFV of at least one SEV subassembly vwith v being an integer between 1 and V, V being greater than or equal to 1 and less than or equal to H. This grouping is carried out from time sequences of the time-stamped power supply frequency values ​​of the respective equipment during the period T. Preferably, the power supply frequency values ​​of a piece of equipment are taken at regular time intervals. This time interval is the same or shorter than that of the elementary duration. These time sequences are, for example, supplied by the control unit UCO to the calculation unit UCA. The grouping REG3 of equipment with variable power supply frequency EQ h is carried out according to a criterion of grouping of equipment with variable power supply frequency EQ h so that the GFV group is composed of at least one SEV subset v . Each SEV subset v of the GFV group includes at least one EQ equipment hof the EV set of variable power frequency equipment EQ h with h= 1 to H. In one embodiment, the criterion for grouping equipment with variable power supply frequency is a correlation criterion between the time sequences of power supply frequency of the EQ equipment h of the set EV for the entire duration of the set of elementary periods T k or period T. Alternatively, the EPIND2 estimation step is devoid of a step for grouping equipment with variable supply frequency. In this case, each SEV subset v from the GFV group is one of the EQ variable power frequency equipment h. The number V of subsets SE vin the GFV group is then equal to H. The estimation step EPIND2 also differs from the EPIND step in that it includes a step of adding AJ the GFV group to the G group resulting from the grouping step REG2 so as to update the G group. After the adding step AJ, the GFV group is part of the set E. Thus, the G group includes each subset of the GFV group, the G1 subgroup and the G2 subgroup. J is therefore the sum of the number of subsets in the GFV group, in the G1 subgroup and in the G2 subgroup. Each subset of the GFV group is one of the subsets of the G group. Each of the GFV group, and of the G1 and G2 subgroups can be empty or non-empty. Step EPIND2 also differs from step EPIND in that step DNB2 differs from step DNB in ​​that it includes, in addition to the steps of step DNB, a step of determining the number of SEV subsets. vof the GFV group and comparing this number to zero. Just as in the example of Figure 3, the estimation step EPIND2 includes a step of estimating a power consumed by at least subsets of the group G from the total elementary consumptions CET(k) and the individual elementary operating times of the subsets SE j, with j= 1 to J of group G. Step EPIND2 further differs from step EPIND in that the parametric model is a linear regression model only when each of the first subgroup G1 and the group GFV is empty and in that when at least one of the first subgroup G1 and the group GFV is non-empty, step EPIND2 comprises a step of estimating a power consumed by a subset of group G from the elementary total consumptions CET(k), the elementary individual operating times of the subsets SE j,with j= 1 to J of group G during the respective elementary periods T k and time sequences of power supply frequencies of the SE subassemblies v of GFV group equipment. The time sequence of the power supply frequency of an SEV sub-assembly v is for example the time sequence of the power supply frequency of a piece of equipment in the subset or the time sequence of the averages of the power supply frequencies of the equipment in this subset. We model the average power consumed ^^^^ ℎ ( ^^^^) of each SEV equipment subset v for v= 1 to V at frequency F by a polynomial of the type: z is an integer greater than or equal to 1, Z is the degree of the polynomial. For each subset of equipment SEv with variable supply frequency, at each frequency F, there is a corresponding theoretical average power consumption. However, the power actually consumed is different from the theoretical power consumption. It depends, for example, on parameters. For example, the power consumed by a given equipment at a given power may depend on the temperature. We therefore model the distribution of the power consumed by each subset SEVv at a given frequency F by a normal distribution: ^^^^ ^^^^is the standard deviation of the normal distribution. In other words, the distribution of individual elementary power values ​​Pv(k) of each SEVv subset with variable supply frequency is modeled by a Gaussian distribution whose mean is a polynomial model of the supply frequency. This model is defined by the following parameters: ^^^^ ^^^^ and, ^^^^ ^^^^0 , ^^^^ ^^^^ ^^^^for z = 1 to Z. Z is predefined. The distribution of the individual elementary power values ​​Pf(k) of each subset SEFf, i.e. of each subset of the subgroup G1 and of the subgroup G2, is modeled by a unimodal or multimodal distribution as described with reference to FIG. 3 for each subset of the group G. Thus, when at least one group taken from the subgroup G1 and the group GFV is non-empty, the method comprises a step AL2 of implementing a stochastic approximation of expectation-maximization algorithm also called SAEM, acronym for the English expression “Stochastic Approximation of Expectation-Maximization” to determine the power consumed by each of the subsets of the group G. This step AL2 differs from the step AL previously described only by the set θN of parameters which differs by the set of parameters in that the parameters associated with each of the subsets SEV vof variable power frequency equipment, of the first set of parameters modeling the distributions of the values ​​of the individual elementary powers, are as follows ^^^^ ^^^^ et, ^^^^ ^^^^0 , ^^^^ ^^^^ ^^^^ pour z = 1 à Z. The parameters associated with the subsets of the G1 and G2 subgroups are as described previously. The parameters associated with each of the SEV subsets vof variable supply frequency equipment are, for example, randomly initialized during the initialization step IN. Alternatively, the method is devoid of the DNB verification step and comprises the implementation of the REGL algorithm and the ESTE step. Hardware From a hardware point of view, the UCA calculation unit can be seen as a calculator interacting with a computer program product. The calculation unit comprises at least one computer, for example, a microcomputer, a computer network, an electronic component, a tablet, a Smartphone or a personal digital assistant (PDA). The calculation unit comprises a memory assembly in which the computer program is stored. The memory assembly comprises, for example, a computer-readable medium.The computer-readable medium is a tangible device readable by the reader of the computing unit, capable of storing electronic instructions and of being coupled to communication means of the computing unit. In other words, the computer-readable medium is a tangible medium. In other words, it is not a transient signal in itself, such as radio waves or other freely propagating electromagnetic waves, such as light pulses or electronic signals. Such a computer-readable storage medium is, for example, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device or any combination thereof.For example, the readable medium is an optical disk, a magneto-optical disk, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a random access memory (RAM), a magnetic card or an optical card. The memory assembly can include an operating system and load the programs according to the invention. It includes registers adapted to record parameter variables created and modified during the execution of the aforementioned programs. A computer program comprising software instructions is then stored on the readable medium.The computing unit comprises a data processing unit comprising at least one electronic circuit adapted to manipulate and / or transform data represented by electronic or physical quantities in registers of the evaluation system and / or memories into other similar data corresponding to physical data in the memories of registers or other types of display devices, transmission devices or storage devices. The data processing unit comprises, for example, memories, for storing data, operatively coupled to the data processing circuit and a reader adapted to read the computer-readable medium.Each function of the computing unit is performed by causing the data processing unit to read a predetermined program from hardware such as the memory array such that the data processing circuitry performs calculations, controls communications, and reads and / or writes data to the memory array. The method is performed on a single computer or on a distributed system among multiple computers (including through the use of cloud computing).The data processing unit comprises at least one of the elements listed below: a set of one or more processors (for example a central processing unit (CPU), a graphics processing unit (GPU), a microcontroller and a digital signal processor (DSP) capable of interpreting instructions in the form of a computer program a programmable logic circuit, such as an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a programmable logic device (PLD) programmable logic arrays (PLA), a system-on-chip (SOC), and / or an electronic card in which steps of the method according to the invention are implemented in hardware elements. The program product may comprise the computer-readable recording medium. Alternatively, the program instructions are obtained from an external source and downloaded via a network. This is particularly the case for applications.In this case, the computer program product comprises a computer-readable data carrier on which the program instructions are stored or a data carrier signal on which the program instructions are encoded. The invention relates to a computer program product comprising the computer-readable medium containing instructions which, when executed by the processing circuit, cause the computing unit UCA to implement the steps of the method according to the invention. The form of the program instructions is, for example, a source code form, a computer-executable form or any intermediate form between a source code and a computer-executable form, such as the form resulting from the conversion of the source code via an interpreter, an assembler, a compiler, a linker or a locator.Alternatively, the program instructions are microcode, firmware instructions, state definition data, integrated circuit configuration data (e.g., VHDL), or object code. The program instructions are written in any combination of one or more programming languages, e.g., an object-oriented programming language (C++, JAVA, Python), a procedural programming language (e.g., C language). The communication means enable communication between the elements of the computing unit and possibly between at least one element of the computing unit and a device external to this computing unit, e.g., the control unit UCO.The communication means can establish a physical link between elements of the system and / or between an element of the system and a device external to the system and / or a remote communication link (wireless) between elements of the system and / or between an element of the system and a device external to the system. The communication means advantageously allow the transmission of data between the control unit UCO and / or the electricity meter CE and the calculation unit UCA.

Claims

CLAIMS 1. Method for estimating an electrical power consumed by a sub-assembly (SE j ) of a group (G) comprising at least one subset (SE j ), each subset (SE j ) of group (G) comprising at least one electrical equipment of a set (E) of electrical equipment, during a period (T) comprising a plurality of elementary periods (T k ), the method being implemented by computer and comprising: - estimating (EPIND) the electrical power consumed by the subassembly (SE j ) using: - individual elementary operating times (a j (k)) of each subset (SE j ) of group (G) during the elementary periods (T k ) respective and, - the total elementary energy consumptions (CET(k)) of the set (E) during the elementary periods (T k) respective.

2. Method according to the preceding claim, comprising: - determining the individual elementary operating durations (a j (k)) of each subset (SE j ) of group (G) during the elementary periods (T k ) respective, - determine the total elementary energy consumption (CET(k)) of the set (E) during the elementary periods (T k ) respective.

3. Method according to any one of the preceding claims, in which estimating (EPIND) the electrical power (P j (k)) includes: - determining, from the total elementary energy consumption and the individual elementary operating times (a j (k)), a set of at least one parameter of a parametric model modeling: - a distribution, over the elementary periods (T k ), of values ​​of individual elementary powers consumed, during the elementary periods (T k ) respective, of each subset (SE j ) of the group (G), - for each elementary period (T k ), the total elementary energy consumption during the elementary period (T k ), by a sum of a possible constant (CST) and the individual elementary energy consumption (CE j (k)) of each subset (SE j ) of group (G) during the elementary period (T k ).

4. Method according to the preceding claim, in which at least when a number of activation / deactivation of at least one subset (SE j ) of the group (G) during the set of elementary periods (T k) is greater than a main threshold, and possibly only when a group of at least one subset of variable supply frequency equipment of the group (G) is non-empty, the model is a probabilistic graphical model.

5. Method according to claim 4, in which the parametric model models values ​​of the individual elementary powers consumed, during the elementary periods (T k ) respective: - of each subset of a first subgroup (G1) of the group (G) by a multimodal distribution, and / or - of each subset of a second subgroup (G2) of the group (G) by a unimodal distribution, - and possibly of each subset of at least one piece of equipment with variable supply frequency of the group (G) by a Gaussian distribution whose mean is a polynomial model of a supply frequency of the subset.

6. Method according to claim 5, comprising: - grouping (REG) subsets (SEj ) of the group (G) so that the first subgroup (G1) consists of each subset of the group (G) having a number of activation / deactivation during the set of elementary periods (T k ) greater than a predetermined main threshold and the second subgroup (G2) consists of each subset presenting a number of activation / deactivation during the set of elementary periods (T k ) less than or equal to the main threshold.

7. Method according to claim 5, comprising: - determining a number of subsets (SE j ) of the first subgroup (G1) so that the model is a probabilistic graphical model only when the first subgroup (G1) is non-empty.

8. Method according to any one of claims 3 to 7, wherein, at least when no subset of the group (G) has an activation / deactivation number during the set of elementary periods (T k) greater than a predetermined main threshold, the parametric model is a linear model.

9. Method according to claim 8, comprising: - determining a number of subsets (SE j ) presenting a number of activation / deactivation during the set of elementary periods (T k ) greater than the main threshold, so that the parametric model is linear only when no subset of the group (G) has a number of activations / deactivations, during the set of elementary periods (T k ), greater than the main threshold.

10. Method according to any one of the preceding claims, comprising a step of prior grouping (REG1) of the equipment of the set (E) according to a correlation criterion between states of the equipment (EQ i ), using status information of the respective equipment, so as to form said group (G) of at least one subset (SE j).

11. Method according to any one of the preceding claims, comprising an initial estimation step (EST1) of the energy consumption of another set of at least one other piece of equipment from a nominal power of the other piece of equipment and status information of the other piece of equipment.

12. Method according to the preceding claim, comprising an initial verification step (VERI) implemented for each equipment of a global set (EG), consisting of verifying whether the equipment complies with a set of at least one importance criterion (C1a, C1b) using a nominal power (P i) predetermined of the equipment and / or equipment status information, the other set of at least one other equipment being made up of each equipment of the set (E) not respecting the set of at least one importance criterion.

13. Method according to any one of the preceding claims comprising, prior to the estimation step, a step of identifying each equipment of the global set (EG) having a total individual operating duration less than or equal to a predetermined maximum portion or greater than or equal to a predetermined minimum portion of the duration T, and a step of removing each equipment identified during the identification step from the global set (EG). 14.A computer program product stored in a non-transitory computer-readable medium, said computer program product comprising code instructions which, when the program is executed by a computer, cause the computer to execute the method according to any one of the preceding claims.

15. A data processing device comprising a processor configured to execute a method according to any one of claims 1 to 13.

16. An industrial system comprising: - a set (E) of equipment (EQ. i ) electrical comprising a group (G) comprising at least one sub-assembly (SE j ), each subset (SE j ) of group (G) comprising at least one electrical equipment of the set (E) of electrical equipment, - an electric meter (CE), - a control unit (UCO) of the electrical equipment capable of activating and deactivating each piece of equipment of the set (E), - a calculation unit (UCA) configured to implement the method according to any one of claims 1 to 13, to estimate an electrical power consumed by a subset (SEj) of the group (G), the total elementary energy consumptions of the individual elementary operating durations (a j (k)) of each subset (SE j ) of group (G) during the elementary periods (T k ) being derived from energy measurements delivered by the electricity meter (CE) and the individual elementary operating times (a j (k)) of each subset (SE j ) of group (G) during the elementary periods (T k ) respective ones coming from the control unit (UCO).

Images